CASSANDRA-21129: Offline TCM dump tool

[dracarys09]

When a Cassandra node fails to start due to Transactional Cluster Metadata (TCM/CEP-21) corruption or issues, operators need a way to inspect the cluster metadata state offline without starting the node. The existing tools (nodetool, cqlsh) require a running node, leaving operators blind when debugging startup failures.

With CEP-21 (Transactional Cluster Metadata), cluster metadata is stored in system tables:

• system.local_metadata_log - Contains transformation entries (epoch -> transformation)
• system.metadata_snapshots - Contains periodic snapshots of ClusterMetadata

When a node fails to start due to TCM corruption or inconsistencies, operators have no way to inspect the metadata state without a running node. This tool fills that gap by reading directly from SSTables.

Thanks for sending a pull request! Here are some tips if you're new here:

• Ensure you have added or run the appropriate tests for your PR.
• Be sure to keep the PR description updated to reflect all changes.
• Write your PR title to summarize what this PR proposes.
• If possible, provide a concise example to reproduce the issue for a faster review.
• Read our contributor guidelines
• If you're making a documentation change, see our guide to documentation contribution

Commit messages should follow the following format:

<One sentence description, usually Jira title or CHANGES.txt summary>

<Optional lengthier description (context on patch)>

patch by <Authors>; reviewed by <Reviewers> for CASSANDRA-#####

Co-authored-by: Name1 <email1>
Co-authored-by: Name2 <email2>

The Cassandra Jira

[krummas]

So this is an emergency recovery tool, hopefully extremely rarely used by an operator, I think we can slim it down a lot, these are the features I think we need here:

• dump metadata to current (or user provided) epoch
  • serialized binary format
  • metadata.toString, to avoid locking us in to any formats
• dump log (with start/end epoch), just toString each entry
• maybe add option to dump system_clustermetadata.distributed_metadata_log if this is run on a CMS node

issues;

• shell script should live in tools/bin/ directory
• tool name - this does not dump sstable metadata, it dumps cluster metadata from sstables, sstable metadata is something different (see tools/bin/sstablemetadata)
• it copies the sstables to $CASSANDRA_HOME/data (or, if that is unset, in to the current directory) - we should create a temporary directory for import and clean that directory up after dumping the metadata, we need something like

                Path p = Files.createTempDirectory("dumptcmlog");
                DatabaseDescriptor.getRawConfig().data_file_directories = new String[] {p.resolve("data").toString()};
                DatabaseDescriptor.getRawConfig().commitlog_directory = p.resolve("commitlog").toString();
                DatabaseDescriptor.getRawConfig().accord.journal_directory = p.resolve("accord_journal").toString();
                DatabaseDescriptor.getRawConfig().hints_directory = p.resolve("hints").toString();
                DatabaseDescriptor.getRawConfig().saved_caches_directory = p.resolve("saved_caches").toString();

to make sure we only touch the tmp directory